Bright Sunshine readings - technical notes

Current sunshine analysis image


Data source

This data is measured by a Davis 6450 solar sensor attached to a wireless Vantage Pro (1) weather station and continuously logged by a standard Weatherlink data logger and software installation. Custom software (XCharter423) is reading the data files automatically and updating the website image at regular intervals (currently 10-15 minutes, but should increase to every 5 mins in the near future).


Chart Key

The chart shows the time-course of global solar irradiance through each successive day and classifies each interval's sunshine level as bright sunshine or not. This chart shows four distinct parameters overlaid:

  1. The main very light-yellow envelope with a grey border shows the time-course of the maximum theoretical brightness levels that could be achieved through each day. In practice, actual readings rarely exceed 90% of the theoretical level and 80% is a more typical maximum measured level (because even on very sunny days there is often a slight milkiness to a blue sky plus haze or pollution that absorbs or scatters a degree of the direct irradiance).
  2. The brown line shows the time-course through each day of the threshold intensity above which sunshine is deemed to be 'Bright Sunshine'. Values measured above this threshold contribute to the BS Total for the day and conversely values below the threshold are excluded.
  3. The consecutive series of bars that appears as each day progresses is the actual set of intensity values measured for the day so far. These are colour-coded so that a brown bar denotes a sunshine level below the BS threshold while a yellow bar shows a bright value, ie included within the Bright Sunshine total.
  4. The grey areas under the maximum envelope shortly after sunrise and before sunset indicate time periods that are excluded from contributing to the BS total even if a measured intensity value exceeds the BS threshold. The reason for these is to help maintain comparability with historic records of Bright Sunshine that would have been measured by a Campbell-Stokes recorder. The CS recorder is generally reckoned to be incapable of burning a trace at low sun angles even when the sky is clear. The exact criterion for the cut-off periods is debatable. Some people suggest that CS traces typically don't register within an hour of sunrise/sunset; others that a sun elevation angle of at least 5° is needed; others that an intensity of at least 100W/sqm needs to be present. These are of course all related expressions of roughly the same process and we can debate which one might be closer to the truth. But the principle of having some cut-off period to maintain comparability with old CS data seems sound enough.
Site Exposure

The test site currently being used to collect this sunshine data has significant horizon limitations. The sensor can see the sun clearly in summer from about 08:30 to around 18:00, but outside of this time period, trees in the vicinity start to obscure the horizon and so plotted values before 08:30 and after 18:00 will show artefactually low values. This exposure is tolerable for current test purposes but obviously the BS totals are not genuine full figures and shouldn't be interpreted as such. At some point in the near future it is hoped to start collecting data at a site with more open exposure (though finding a site that is available and secure and that offers an unfettered view of the horizon throughout a midsummer day is no easy matter!).

Measurement Intervals

The time interval over which sunshine intensity is measured obviously has some influence on BS evaluation. To start with, the Davis VP solar sensor updates its reading roughly every minute, so this places a minimum resolution or granularity on the brightness data. In other words, whatever other data manipulations might be done, there is no way of getting more time resolution in the sunshine data than once per minute updates provide. Then, in the data feed used in these particular analyses, we're reading the Weatherlink archive data files and this imposes a further resolution limit, depending on the archive interval currently set in the active data logger. While this can also be at a 1-minute frequency, in practice such a high sampling rate generates a large body of data with considerable redundancy. A more typical configuration for routine weather monitoring might use an interval of 5 or 10 minutes instead and the single brightness value reported for each interval will be the mean value over the full interval.

While an interval of 5 or 10 minutes is frequent enough for most weather parameters, it does inevitably lead to some averaging of brightness levels because the sun can obviously go behind a cloud midway through a logging interval. The value reported will then be the mean of the brighter and the duller intensity levels during the interval. This averaging tendency will of course be more exaggerated with a 10-minute interval set than over a 5-minute one. We are currently using a 10-minute interval for development of the initial graphics here but will switch to 5 minutes in the near future. Note though that the choice of interval not only affects archive data volumes but has consequences for the CPU power required and also on the details of the chart design - which will be discussed on another page - so there are several issues to consider when setting the interval and this is why an immediate jump to a 5- or even a 1-minute interval was not made from the start of this study.

This averaging of the brightness over several minutes tends to mean that on days with frequent sunny intervals interspersed with cloudy conditions the typical brightness value in an interval will be reduced (ie under such weather conditions several of the intervals logged will represent a mixture of sunshine and cloud). However, this effect need not cause an under-recording of sunshine hours.

In the process of validating this general approach of calculating BS hours from total brightness data, we've needed to calibrate our algorithm against BS totals measured simultaneously but independently with a CS recorder. By correlating data over periods of eg a month and by including both winter and summer months, we're able to derive an overall mean brightness threshold for BS assignment that gives optimum agreement with the CS-measured totals. This does mean that the BS threshold is somewhat lower than it would be if we were able to measure very short sunshine intervals because of the need to bring into play measured brightness intervals that would have been a mixture of sunshine and clouds and this does also make it difficult to assign any individual interval with a value around the threshold as categorically either BS or not. But, overall, across a month's data - which is after all the main concern for most weather observers - agreement between calculated and measured BS hours can be good, eg ±10% or better.